The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several shell commands using the claude CLI (gateway stop, plugins install, config set, gateway install) and standard utilities like grep and tail to manage the agent's environment and monitor logs.
[EXTERNAL_DOWNLOADS]: The command claude plugins install manifest facilitates the download and installation of external plugin code from the platform's plugin registry.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data which presents an injection surface.
Ingestion points: User-provided API keys, custom endpoints, and the contents of the ~/.claude/logs/gateway.log file in SKILL.md.
Boundary markers: The skill does not use explicit delimiters when interpolating user input into commands.
Capability inventory: The skill has the ability to execute shell commands and read local log files.
Sanitization: The skill implements a basic format check requiring the mnfst_ prefix for API keys before they are used in configuration commands.

manifest